Most computer applications (hereinafter “applications”) are very complex systems that, due to their complexity, require significant testing to ensure that the application will execute as desired.
To facilitate the testing of applications, test cases or test suites (essentially a collection of test cases) are designed, implemented and used to test a portion or the whole of an application (often referred to as the subject under test). In many applications, these test cases manipulate the external facade or interface of the subject under test. The results of these test cases are then analyzed and evaluated. As many applications are quite complex, several, sometimes hundreds, of test cases are used for the testing of a single application.
For example, a database application may need to be tested to determine whether data can be added to the database (this is the test case). A test script would need to be created to implement the test case. The exemplary test script could include several steps, instructions or processes to test this aspect of the application including: gaining access to update the database, transmitting the update request; receiving confirmation that the update request has been executed; reading from the database to determine if the data in the update request was stored successfully; and then logging off from the database.
Presently, testing enables a tester to insert a verification point at all, some or none of the steps that are performed in the test script that implements a test case. A verification point, when inserted into the test script, will issue or output a binary value (usually a single bit of data—e.g., a boolean value) that indicates whether a step(s) with which the verification point is associated was successfully or unsuccessfully completed. The output of the test script execution, which includes the values output by the verification points, is typically stored in a test case execution script log.
The verification points enable testers to analyze the test case execution log to ascertain which processes in the test case failed and, thus, which portions of the subject under test need to be investigated to rectify any problems (i.e., solve any bugs in the application, if necessary).
Some testing tools presently on the market will terminate the execution of a test script as soon as a single verification point is reached which indicates that an associated step was not performed successfully. Other testing tools will attempt, if possible, to continue processing the test script execution and outputting data from the verification points.
As a result of the data output by the verification points, a test script execution will generate an output which indicates whether the application successfully or unsuccessfully performed the test case (the test output). The test output of the test script execution (which is in addition to test logs that are typically generated) is, like the output of a verification point, a binary value. If even a single step in the test script failed then the entire test script execution will output a value also indicating failure of the execution of test script (i.e., the application failed in performing the test case). Only when all steps are successfully performed will the test script output indicate successful completion.
It will be appreciated that when a deadline for release of an important application draws near, the application developers (the development team) are under considerable stress to have the application ready for release with a minimum number of bugs or quality or stability related issues. At this time, review of the test output data generated by the test scripts executions is often at its highest. During this review, much effort and time is spent attempting to ascertain the reason certain test scripts execution have failed to complete successfully indicating a likely problem with the subject under test. It is common in the industry, especially as deadlines loom, to disregard the test results from some test scripts execution for a variety of reasons. For example, a test script may be considered by the development team as testing an “unimportant” or “inconsequential” aspect of the subject under test. Other test scripts may be disregarded on an ad hoc basis because of a developer's intuition or experience indicating that the unsuccessful result from the execution of a test script was not as a result of a failure of the subject under test but rather another application or component (e.g., the operating system, memory management, etc.). Unfortunately, this analysis is often inaccurate or misguided and potentially results in the release of an unacceptable application. Moreover, the rationale for ignoring some test results while relying on others is not documented which may prove unfortunate after a product has been released.
In a further shortcoming of known systems, significant time may be spent analyzing the unsuccessful results issued by a test script execution in an attempt to ascertain the cause(s) of the failure. This analysis is premised on the possibility that all failures should be determined and that most or sometimes all of the failures are the result of flaws or bugs in the subject under test. However, in some instances, the failure may be the result of a failure in another portion of the computer system(s) on which the subject under test is relying. In such an instance, a developer or tester may have misspent significant time and energy investigating a flaw which is not caused by the subject under test.
For example, returning to the database example described above, suppose that the test script process for logging off from the database required transmitting to the database a “logoff” command. Further assume that the database itself (and not the application/subject under test) had a slight bug that issued an error message despite properly logging off a user. In this scenario, the verification point associated with the portion of the test script for logging off would output indication of a failure and the test script, even if every other process completed successfully, would, as a result of the reported logoff failure, also generate test output indicating a failure of the subject under test. Moreover, unfortunately while this failing result may be technically accurate (i.e., the test script did not successfully complete each of the processes contained therein) it is misleading and inaccurate in substance (i.e., the subject under test actually successfully performed all of its processes and only another element relied upon by the subject under test failed). This situation could lead to significant time and energy being spent by a developer attempting to eradicate a bug in the subject under test which does not exist (the bug exists in the database).
Accordingly, a testing measurement which addresses, at least in part, some of the shortcomings described above is desired.